I am trying to run my model under myopic foresight but I am having several conflicts showing (specially when the accumulated activity is bounded and annual activity isn't).
The problem is that the LST file shows an upper bound which is 10 times smaller than the upper bound I imposed for the process whose equations are in conflict (I verified in the process master data and the the upper bound I imposed is the one shown).
I attach two screenshots : one of the conflict shown is the LST file and the other from the process master.
07-04-2020, 03:36 PM (This post was last modified: 07-04-2020, 03:37 PM by Antti-L.)
Because cumulative amounts are typically quite large, cumulative bound equations are scaled down by 10. Both the LHS and the RHS sides are of course scaled. So, when the original constraint would be:
CumAct(r,p) ≤ ACT_CUM(r,p,'UP')
The scaled constraint is then simply:
CumAct(r,p)/10 ≤ ACT_CUM(r,p,'UP')/10
Thus, as far as I can tell, no conflicts can be arising from that scaling. However, I can see from the Cplex report that your activity bounds appear to be conflicting the cumulative bound of 5.25 you have specified, would you not agree?
yes, I totally agree with you regarding the conflict between the activity bound and the cumulative bound on activity. I have to limit the annual activity of the process to avoid this conflict.
I think that my problems is initiated by something else. In fact, I thought that the activity bounds were conflicting the cumulative bound 5.25 I imposed because I am under myopic foresight (with 20 years for each window and 10 years overlapping between periods). Thus, I decided to add a bound on the annual activity in the region in which the conflict occurs (cf the attached file). But I had the same message from Cplex. My model is infeasible because of the conflict between the activity bound and the cumulative activity. I changed the I/E rule from 15 to 5 even if I wasn't convinced this could have an effect on the results since rules 5 and 15 are quite similar apart from the migration rule.
The message from Cplex stays the same and the bound on annual activity wasn't taken into account. I am not having any UC on activity that could conflict the ATCBND set to 0.52 /year.
Assuming the period length is 1 for 2010 and 3 for 2016, the total lower bound is already 5.248494, while your upper bound is shown to be 5.24857. Thus, there would be room only for 0.00008 of additional cumulative activity. But you have also additional (small) lower bounds in excess of that amount, and so it seems clear that you get an infeasibility.
I am sorry but I am not able to see any details of your model, and so I don't know why you have all those lower bounds in your model. But they don't seem to be fixed bounds, because Cplex reports them as lower bounds. Therefore, they don't seem to be resulting from any previous steps of the time-stepped algorithm, and neither from fixing initial periods to previous solution, because those would be fixed bounds.
Maybe you can explain yourself why you have all those lower bounds on the activity, which together cause the conflict?
